Orkes Conductor - AI Agent Platform
Overview
Design a platform that allows users to easily create, customize, and deploy AI agents without needing technical expertise.
Role: Lead UX Designer
Timeline: 10 weeks
Team: 1 PM, 2 engineers, 1 UX designer
Timeline: 10 weeks
Team: 1 PM, 2 engineers, 1 UX designer
Problem
How might we create a platform that makes building and managing AI agents intuitive for all users—while keeping the flexibility advanced users expect?
Research & Insights
User interviews (10 participants): Small business owners, developers, and customer support managers.
Key Findings
- Users want a “wizard-like” setup experience.
- Most feel overwhelmed by terminology like “API key” or “prompt chain.”
- Customization should feel visual, not code-based.
Opportunity
Make AI configuration visual and conversational—so users feel like they’re building with AI, not programming it.
Solution = a drag-and-drop AI agent builder with conversational setup.
Users can define an agent’s:
- Purpose (via guided prompts)
- Knowledge sources (files, URLs, databases)
- Personality & tone
- Integrations (Slack, email, web chat)
Key UX Features
- Onboarding Wizard: Step-by-step guide that uses plain language (“What do you want your agent to do?”).
- Visual Flow Editor: Nodes for actions, data sources, and responses—no code required.
- Preview Chat: Real-time testing area to validate agent behavior.
- Dashboard: Track usage, performance, and feedback loops.
Design Process
- Low-fidelity sketches explored different metaphors (chat setup vs flowchart).
- Prototype testing revealed users preferred a conversational setup followed by visual refinement.
- Iterated on information hierarchy and terminology simplification—replacing “LLM” with “Language model,” etc.
Outcome & Impact
Usability score: 92% task completion in prototype tests.
Average build time: Reduced from 45 minutes (in competitor tools) to under 10 minutes.
Beta feedback: “It feels like the AI is helping me build itself.”
Next Steps
Integrate analytics to improve agent performance recommendations and enable multi-agent collaboration features.
Key Takeaway
Empowering users to build AI agents isn’t just about simplifying interfaces—it’s about designing trust, clarity, and collaboration between human and machine.
Browse the Prototype
(Enter fullscreen for best viewing experience)
